# Eighteen runnable research utilities

The [sixteen earlier utilities](PROTOTYPES-2026-10-09-v5.md) now include two computational-model references. Seventeen saved component reports contain **2,056 passed checks**: the previous 1,745 plus 137 movement, 131 transport, 12 new documentary-contract and 31 report-input archive checks. No source proof or formal program verification was executed.

## Exact finite movement comparison

[kserver_reference.py](tools/kserver_reference.py) accepts a rational metric, initial labeled tuple and fixed request word. It computes an attained offline optimum by conventional DP, replays a witness, and evaluates fixed-label, nearest-label and independent uniform-label policies exactly. It implements no source competitive policy. It rejects nonmetrics and caps eight points, four servers, 64 requests and three million state-label transitions. Budget exhaustion leaves the complete comparison unknown. Dimensions bound frontier/backpointer storage; work units exclude input validation and do not represent bit operations or wall time.

```sh
python3 research/tools/kserver_reference.py research/fixtures/movement-startup-2026-10-09-v1.json
```

[137 checks](snapshots/2026-10-08-baseline/movement-reference-validation-2026-10-09-v1.json) compare complete short label-string enumeration with DP and policy expectations, then check witness replay, horizon distinctions, metrics and budgets. The saved 30-request startup fixture has fixed-label cost 30 and offline optimum 2. This is a small finite control for the source's activation/additive-cost distinction, not performance evidence on a customer workload.

## Exact transport sharpness geometry

[transport_sharpness_reference.py](tools/transport_sharpness_reference.py) reproduces the source's uniform-square three-atom example. It clips rational affine halfplanes, integrates nine map-cell intersections, and compares masses and map norm with analytic formulas. Target W1/W2 use the explicit matching and second-coordinate lower witness; no general transport optimizer runs. It accepts a 256-bit rational parameter 0<a<1/2 and caps 10,000 polygon-vertex examinations. Partial work returns unknown. It proves no arbitrary transport-map or solver guarantee.

```sh
python3 research/tools/transport_sharpness_reference.py research/fixtures/transport-three-atom-2026-10-09-v1.json
```

[131 checks](snapshots/2026-10-08-baseline/transport-reference-validation-2026-10-09-v1.json) cover rational geometry, target metrics, exact half-exponent growth and input/budget boundaries. [Saved examples](snapshots/2026-10-08-baseline/computational-model-examples-2026-10-09-v1.json) include movement comparisons, explicitly conditional activation schedules and transport controls. No source constructor time is measured.

## Documentary and semantic records

The [thirteen-contract registry](prototypes/curated-contracts-2026-10-09-v3.json) adds language-class, uniform k-server and Brenier-map contracts, with [12 checks](snapshots/2026-10-08-baseline/semantic-contract-validation-2026-10-09-v1.json). Complete documentation still grants no applicability/independent-verification certificate and does not verify supplied evidence contents. The [new joined report](prototypes/computational-model-evidence-bundle-2026-10-09-v2/report.md) flags eleven gaps in one self-authored startup claim. It archives all six exact report inputs. Earlier bundles also have separately recovered input archives whose bytes match the original recorded hashes; their original manifests/reports remain preserved. [31 archive checks](snapshots/2026-10-08-baseline/evidence-archive-validation-2026-10-09-v1.json) verify those bytes and historical registry counts without executing the archived tools.

[41 selected definition observations](snapshots/2026-10-08-baseline/definition-hole-review-queue-2026-10-09-v2.json) have no detected intended-model mismatch in this limited review. Seven interfaces/24 entries remain pending. This is source inspection, not a trusted checker or whole-proof result. [Operational limits](FEASIBILITY-2026-10-09-v5.md).

## Full utility directory

| Utility | Capability |
| --- | --- |
| [Proof preflight](tools/proof_preflight.py) | Static module/config/trust metadata |
| [Withdrawal impact](tools/withdrawal_impact.py) | Explicit corrections/dependency paths |
| [Embedding audit](tools/audit_embedding.py) | Finite floating-point distortion |
| [GAD capacity objective](tools/gad_capacity.py) | Finite scalar evaluation |
| [Claim contract audit](tools/audit_claim_contract.py) | Thirteen curated documentary contracts |
| [Periodic interface reference](tools/periodic_interface_reference.py) | Cubic-periodic perimeter formula |
| [Binary waveform audit](tools/audit_binary_waveform.py) | Exact correlations/sample spectrum |
| [Ramanujan graph audit](tools/audit_ramanujan_graph.py) | Exact strict finite spectral bound |
| [Cyclic-chain reference](tools/cyclic_chain_reference.py) | Small-group shortest cyclic quotient chain |
| [Evidence bundle](tools/evidence_bundle.py) | Source fingerprints/joined report |
| [Contingency reference](tools/contingency_reference.py) | Exact bounded table count/rank |
| [Matching certificate audit](tools/audit_matching_certificate.py) | Feasibility/attaining cardinality bound |
| [Existing-solver adapter](tools/solve_and_audit_matching.py) | Bounded proposal/separate audit |
| [Perfect-pairing reference](tools/perfect_matching_reference.py) | Exact count/rank/edge fragility |
| [Switch-chain audit](tools/audit_switch_chain.py) | Exact bounded kernel/law/host calibration |
| [Tree thinness audit](tools/audit_tree_thinness.py) | Exact bounded all-cut ratio/certificate |
| [Movement reference](tools/kserver_reference.py) | Exact bounded offline optimum and named-policy expectations |
| [Transport sharpness reference](tools/transport_sharpness_reference.py) | Exact three-atom polygon/mass/map-distance control |

Dated validation drivers/output files preserve their earlier editions. Existing successful output paths cannot be reused. Finite checks, syntax/hash consistency and an elapsed work window do not establish broad algorithm superiority, buyer demand or profitability.